Small Clusters of Large Nodes for Big Data Deployments
In the world of Big Data, scaling out the database is the norm. As a result, Big Data deployments are often awash in clusters of small nodes, which bring with them many hidden costs. For example, scaling out usually comes at the cost of resource efficiency, since it can lead to low resource utilization.
So why is scaling out so common? Scaled out deployment architectures are based on several assumptions that we’ll examine in this white paper. We’ll demonstrate that these assumptions prove unfounded for database infrastructure that is able to take advantage of the rich computing resources available on large nodes. In the process, we’ll show that Scylla is uniquely positioned to leverage the multi-core architecture that modern cloud platforms offer, making it not only possible, but also preferable, to use smaller clusters of large nodes for Big Data deployments.